Artificial intelligence and machine-learning are novel exciting tools which we hope will help drug development. AI is analysis of very large data sets with statistical machine-learning methods. The simple way of explaining the role of AI is that it could be used in drug development to analyse big sets of data in order to identify potential new drugs.
But before we discuss challenges and opportunities that AI could provide we can explore a bit more in the current drug development process. According to the statistics there are 16 new molecular entities launched per year between 1950 and 2014. Slow drug development process and all its challenges cause costs to go up by 8% per year. Some of the potential challenges could be that little is known about the biological mechanisms of the particular disease; the potential drug target is difficult to be reach or result in other complications; some molecules are from classes, which cannot be used as drugs due to various reasons, etc. Another factor is publication bias and patent limitations which prevent scientists, working in drug development to have adequate information about potential molecules. The tendency of pharma companies to outsource parts of drug development process to CROs is another challenge for scientists because it prevents them from access to important information.
While artificial intelligence and big data are bring exiting new opportunities they also have their own risks.
- One of the biggest risks of creating big data base with all information in medical chemistry will be that may end up with `everyone doing the same thing`;
- Identifying new compounds is very complex process;
- AI has been tested mainly on training data so far;
- Lack of interpretability – it does not show what parameters it has taken into account to produce the final result;
- There is no method to assess how adequate are the final results which could potentially result in increased costs;
- There is a risk of algorithmic bias.
All these risks shows that in order to use artificial intelligence you need a large sets of data that could be used to train and establish method to analyse and interpret the final results. However, this should not discourage the usage of AI in drug development because there are great opportunities ahead of it.
Author: Olga Peycheva
Olga is a clinical research professional who has been working in clinical research since 2005. She has extensive experience in clinical research in Eastern and Western Europe.
Originally published on 1 Apr 2019